We're using a blue OSRAM projection LED, for fluorescence neural imaging. Rated for 20W of light with 20A of current drive, and with spec'd 30A pulse, and 40A surge for 10us, we drive it with 200A for 45us to get our needed photons (each $30 LED lasts about 2 months). Straight-forward 200A current-source-drive circuit, AoE x-Chapter writeup, read it here:
The light is focused on a 1 mm area on the fish's brain, and tracks its motion. Interfering pieces of stuff and tissue, momentary angled orientations, etc., are behind the high light level requirement. A continuous map of brain nerve action is made,* and pairs of closely-spaced images were made withing 150us (using two 45us flashes) and processed in real time, to determine the velocity vectors, so they can move the tracking stage to follow the fish and keep the brain in range of the TI mirror.
I added the paper to DropBox, read all about it.
The OSRAM projection LED actually has four sections, they have to select and use only the LEDs that have similar light levels from all four sections.
At this young age, the fish is transparent. The goal is to see exactly which brain neurons are controlling what swimming behavior, in real time. (The zebra fish brain has been fully mapped.) Sheesh! I think I'll stick to electronics. Ahem, it's already fully mapped.
Yes, me too, frequently. But DH, the young postdoc, had already wired up circuit A on his own, and came to me with the poor pulse performance, asking what to do. I thought and gave him a kit of parts with a description, and he made it work the same day. I knew the MOSFET current-sink idea would work, glanced at the datasheet, knew it could handle the task, and skipped the calcs described. The whole thing took 10 minutes, it didn't even get a RIS project number. No opamps allowed (drive Ciss = 7600pF in under a us).
Also transparent to blue. There's a calcium florescence that shows specific neuronal activity, check out the paper.
No side effects on the fish, light duty cycle under 1%. Bright enough for florescence, but dim enough to avoid damage. An individual fish can be tracked for hours.
make sense to me. You stick the dye into the cell, and the fluorescent emis sion - properly detected - tells you about local calcium level.
One of my friends invented an improved confocal microscope, and another of them used one of the improved confocal micrscopes for some kind of fluoresc ence research (thanks to my introduction) but they are both in Melbourne an d I don't see either of them often.
I think that's two months of lab operation, I suppose 10 experiments, maybe 2 to 4 hours each. LEDs cost $28 each. The 200A risetime is one or two us, we haven't experimented with slowing that down. They're very jealous for every little bit of light they can get in the 45us. It's amazing the LEDs survive at all, we're running them 22x the absolute max spec.
BTW, there are two 200A 45us flashes, the 2nd after a 70us delay. The first image is stored in the readout cells of the sensor, the second one in the detector cells, transferred and readout after the first one. These two images, 115us apart, are GPU processed within 1.5 ms, and used to determine fish's direction and velocity, and update the tracking system once every 4ms, to keep the camera stage and 1mm beam centered on the brain. There's a flash pair every 4ms, for a 2.3% duty cycle. Ouch. That's twice as high as I was told before. Given that, I'd say we're at 45x the max spec. Kind of a John Larkin thing, keep turning it up until it breaks, then back off a bit.
Are you sure the two closely spaced pulses 70us apart are not driving the t wo separate halves of that OSRAM projection LED? By processing the differin g intensity of the tracked object returns, it would be a way of determining its angular offset from the mirror pointing angle, a standard technique fo r high resolution tracking. Haven't read the paper, but it otherwise doesn' t make any sense the 4ms track samples give adequate accuracy for the motio n dynamics of the target, but then they need this double tap at 70us spacin g. The close spacing is obviously a time sequential way of getting these tw o offset samples at essentially the same target position. Any way, if that' s what they're doing, then they have two MOSFET drives, putting the duty cy cle back at 1.15% per copy. Of course, I could be completely off about this .
No, only one driver, driving all four sections. In fact Drew selects from the tray of LEDs, the ones with best-matched light outputs from all four sections. The 45us duration is chosen to get minimum image blur. Two images, processed by a massive GPU home-made computer, gives a full 2D velocity vector, with a few ms to spare. So dunno why they picked the close spacing, I'll ask. I think you'll enjoy reading the paper.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.